The aim is to determine if functional magnetic resonance imaging (fMRI) can help predict whether task-oriented physical therapy is likely to enhance outcomes related to walking and whether fMRI can inform clinicians about the optimal duration and intensity of treatment. A standardized therapy called body weight-supported treadmill training (BWSTT) and over ground walking will be given to 32 subjects who, at 6-18 months after hemiparetic stroke, still walk at less than the minimal velocity needed for community ambulation. Subjects are stratified by initial walking speed (<40 cm/sec vs 40-75 cm/sec), which relates to walking limitations and neurological impairment. The study serially quantifies therapy-related changes in walking speed, endurance, gait pattern, and physiological measures relevant to gains in motor control. In parallel, serial fMRI studies are performed using a standardized activation paradigm of passive and voluntary ankle dorsiflexion of the paretic leg. fMRI will quantify changes in synaptic activity within primary and nonprimary cerebral sensorimotor regions and reflect subcortical and spinal plasticity, as well as activity in areas engaged for attention and learning, as better motor control is engrained by the intensive, progressive rehabilitation strategy. For this longitudinal design, subjects continue training in blocks of 12 sessions, until their walking speed reaches a plateau and no further reorganization is found by fMRI performed at the end of each block. We anticipate that gains in motor control for ankle dorsiflexion will be reflected over time of training in quantifiable changes in representational plasticity within the supraspinal constituents of the network for walking. Using statistical modeling, we will examine relationships between walking related gains and the fMRI patterns induced over the course of therapy. Of great interest is whether the initial pattern or the early training-induced evolution of cerebral activity can serve as a physiological assay to predict the ongoing capacity of the spared network to be modulated by more therapy. We will then test a statistically derived predictive model by training a new group of 12 subjects. The study is a step toward using fMRI as a tool to develop theory- and evidencebased practices that lessen disabilities after stroke. [unreadable] [unreadable] [unreadable]